Quantitative evaluation of six multi-spectral, multi-resolution image merger routines

Currently, most image merger routines are judged on the enhanced appearance of the hybrid imagery that they create. Little or no quantitative analysis is given which represents a technique\u27s ability to preserve the radiometry of the hybrid imagery. For this reason, a quantitative analysis of six multi-resolution, multitemporal, multi-sensor, multispectral image merger routines was performed. These techniques were designed to incorporate the high resolution spatial information of a panchromatic image with the spectral information of a low resolution multispectral image. The panchromatic imagery came from the French satellite Systeme Pour l\u27Observation de la Terre (SPOT) and the multispectral imagery came from the NASA Landsat Thematic Mapper (TM). This study served as a quantitative test of three commonly used merger techniques (intensity-hue-saturation (IHS) transformation method, principal component method, and high-pass filter technique) and three merger techniques developed by the Digital Image Processing and Remote Sensing (DIRS) laboratory at RIT (ratio method, extended regression method, and global regression method). These routines were judged on the basis of their radiometric integrity and their dependent and independent classification accuracies. Statistical analysis of these metrics showed that the DIRS global coefficient routine provided the best enhancements of the multispectral imagery for the scenes and testing metrics selected